A study using 51 wheat, 56 barley and 34 oat grain samples was conducted to investigate the feasibility of predicting the apparent metabolizable energy (AME) value of these cereals for poultry. Stepwise regression analyses were performed to evaluate the relationship of AME with starch, ether extract (EE), crude fiber (CF), soluble sugar (SS), ash and crude protein (CP) (for wheat and barley grain samples) or dry matter (DM), CF, ash and CP (for oat grain samples) as independent variables. According to the stepwise regression analyses, SS, CF and ash for wheat, CF, EE and starch for barley and CF and CP for oat were found to be useful predictors for AME prediction. Also, multiple linear regression (MLR) and artificial neural network (ANN) methods were developed to find the best models which can estimate the AME content of these cereals. Mean square deviation, Mean square variation and their components were used to evaluate the performance of MLR and ANN models. The results showed that AME of wheat can be predicted by SS, CF and ash. The CF, EE and starch are good independent variables to estimate AME content of barley samples. Also, CF and CP are good predictor parameters for AME prediction in oat samples. In case of model performance, the accuracy of the ANN model was stronger than MLR. Based on these results, it was concluded that the use of chemical composition in combination with the ANN model is a promising method to predict AME of wheat, barley and oat grain samples in poultry nutrition. Manuscript profile